How long will my mouse live? Machine learning approaches for prediction of mouse life span

Predictive Models of Life Span in Old Female Mice Based on Immune, Redox, and Behavioral Parameters

The development of mathematical models capable of predicting the lifespan of animals is growing. However, there are no studies that compare the predictive power of different sets of parameters depending on the age of the animals. The aim of the present study is to test whether mathematical models for life span prediction developed in adult female mice based on immune, redox, and behavioral parameters can predict life span in old animals and to develop new models in old mice. For this purpose, 29 variables, including parameters of immune function, redox state, and behavioral ones, were evaluated in old female Swiss mice (80 ± 4 weeks). Life span was registered when they died naturally. Firstly, we observed that the models developed in adults were not able to accurately predict the life span of old mice. Therefore, the immunity (adjusted R2 = 73.6%), redox (adjusted R2 = 46.5%), immunity-redox (adjusted R2 = 96.4%), and behavioral (adjusted R2 = 67.9%) models were developed in old age. Finally, the models were validated in another batch of mice. The developed models in old mice show certain similarities to those in adults but include different immune, redox, and behavioral markers, which highlights the importance of age in the prediction of life span.

Longitudinal Fragility Phenotyping Predicts Lifespan and Age-Associated Morbidity in C57BL/6 and Diversity Outbred Mice

Aging studies in mammalian models often depend on natural lifespan data as a primary outcome. Tools for lifespan prediction could accelerate these studies and reduce the need for veterinary intervention. Here, we leveraged large-scale longitudinal frailty and lifespan data on two genetically distinct mouse cohorts to evaluate noninvasive strategies to predict life expectancy in mice. We applied a modified frailty assessment, the Fragility Index, derived from existing frailty indices with additional deficits selected by veterinarians. We developed an ensemble machine learning classifier to predict imminent mortality (95% proportion of life lived [95PLL]). Our algorithm represented improvement over previous predictive criteria but fell short of the level of reliability that would be needed to make advanced prediction of lifespan and thus accelerate lifespan studies. Highly sensitive and specific frailty-based predictive endpoint criteria for aged mice remain elusive. While frailty-based prediction falls short as a surrogate for lifespan, it did demonstrate significant predictive power and as such must contain information that could be used to inform the conclusion of aging experiments. We propose a frailty-based measure of healthspan as an alternative target for aging research and demonstrate that lifespan and healthspan criteria reveal distinct aspects of aging in mice.

Progress in the study of aging marker criteria in human populations

The use of human aging markers, which are physiological, biochemical and molecular indicators of structural or functional degeneration associated with aging, is the fundamental basis of individualized aging assessments. Identifying methods for selecting markers has become a primary and vital aspect of aging research. However, there is no clear consensus or uniform principle on the criteria for screening aging markers. Therefore, we combine previous research from our center and summarize the criteria for screening aging markers in previous population studies, which are discussed in three aspects: functional perspective, operational implementation perspective and methodological perspective. Finally, an evaluation framework has been established, and the criteria are categorized into three levels based on their importance, which can help assess the extent to which a candidate biomarker may be feasible, valid, and useful for a specific use context.

Intelligent high-throughput intervention testing platform in Daphnia

We present a novel platform for testing the effects of interventions on the life- and healthspan of a short-lived freshwater organism with complex behavior and physiology-the planktonic crustacean Daphnia magna. Within this platform, dozens of complex behavioral features of both routine motion and response to stimuli are continuously quantified over large synchronized cohorts via an automated phenotyping pipeline. We build predictive machine-learning models calibrated using chronological age and extrapolate onto phenotypic age. We further apply the model to estimate the phenotypic age under pharmacological perturbation. Our platform provides a scalable framework for drug screening and characterization in both life-long and instant assays as illustrated using a long-term dose-response profile of metformin and a short-term assay of well-studied substances such as caffeine and alcohol.

Integrated genetic analyses revealed novel human longevity loci and reduced risks of multiple diseases in a cohort study of 15,651 Chinese individuals

There is growing interest in studying the genetic contributions to longevity, but limited relevant genes have been identified. In this study, we performed a genetic association study of longevity in a total of 15,651 Chinese individuals. Novel longevity loci, BMPER (rs17169634; p = 7.91 × 10^-15 ) and TMEM43/XPC (rs1043943; p = 3.59 × 10^-8 ), were identified in a case-control analysis of 11,045 individuals. BRAF (rs1267601; p = 8.33 × 10^-15 ) and BMPER (rs17169634; p = 1.45 × 10^-10 ) were significantly associated with life expectancy in 12,664 individuals who had survival status records. Additional sex-stratified analyses identified sex-specific longevity genes. Notably, sex-differential associations were identified in two linkage disequilibrium blocks in the TOMM40/APOE region, indicating potential differences during meiosis between males and females. Moreover, polygenic risk scores and Mendelian randomization analyses revealed that longevity was genetically causally correlated with reduced risks of multiple diseases, such as type 2 diabetes, cardiovascular diseases, and arthritis. Finally, we incorporated genetic markers, disease status, and lifestyles to classify longevity or not-longevity groups and predict life span. Our predictive models showed good performance (AUC = 0.86 for longevity classification and explained 19.8% variance of life span) and presented a greater predictive efficiency in females than in males. Taken together, our findings not only shed light on the genetic contributions to longevity but also elucidate correlations between diseases and longevity.

Age and life expectancy clocks based on machine learning analysis of mouse frailty

The identification of genes and interventions that slow or reverse aging is hampered by the lack of non-invasive metrics that can predict the life expectancy of pre-clinical models. Frailty Indices (FIs) in mice are composite measures of health that are cost-effective and non-invasive, but whether they can accurately predict health and lifespan is not known. Here, mouse FIs are scored longitudinally until death and machine learning is employed to develop two clocks. A random forest regression is trained on FI components for chronological age to generate the FRIGHT (Frailty Inferred Geriatric Health Timeline) clock, a strong predictor of chronological age. A second model is trained on remaining lifespan to generate the AFRAID (Analysis of Frailty and Death) clock, which accurately predicts life expectancy and the efficacy of a lifespan-extending intervention up to a year in advance. Adoption of these clocks should accelerate the identification of longevity genes and aging interventions.

When will my mouse die? Life span prediction based on immune function, redox and behavioural parameters in female mice at the adult age

The identification of predictive markers of life span would help to unravel the underlying mechanisms influencing ageing and longevity. For this aim, 30 variables including immune functions, inflammatory-oxidative stress state and behavioural characteristics were investigated in ICR-CD1 female mice at the adult age (N = 38). Mice were monitored individually until they died and individual life spans were registered. Multiple linear regression was carried out to construct an Immunity model (adjusted R² = 75.8%) comprising Macrophage chemotaxis and phagocytosis and Lymphoproliferation capacity, a Redox model (adjusted R² = 84.4%) involving Reduced Glutathione and Malondialdehyde concentrations and Glutathione Peroxidase activity and a Behavioural model (adjusted R² = 79.8%) comprising Internal Locomotion and Time spent in open arms indices. In addition, a Combined model (adjusted R² = 92.4%) and an Immunity-Redox model (adjusted R² = 88.7%) were also constructed by combining the above-mentioned selected variables. The models were also cross-validated using two different sets of female mice (N = 30; N = 40). Correlation between predicted and observed life span was 0.849 (P < 0.000) for the Immunity model, 0.691 (P < 0.000) for the Redox, 0.662 (P < 0.000) for the Behavioural and 0.840 (P < 0.000) for the Immunity-Redox model. Thus, these results provide a new perspective on the use of immune function, redox and behavioural markers as prognostic tools in ageing research.

Chronic helminth infection burden differentially affects haematopoietic cell development while ageing selectively impairs adaptive responses to infection

Throughout the lifespan of an individual, the immune system undergoes complex changes while facing novel and chronic infections. Helminths, which infect over one billion people and impose heavy livestock productivity losses, typically cause chronic infections by avoiding and suppressing host immunity. Yet, how age affects immune responses to lifelong parasitic infection is poorly understood. To disentangle the processes involved, we employed supervised statistical learning techniques to identify which factors among haematopoietic stem and progenitor cells (HSPC), and both innate and adaptive responses regulate parasite burdens and how they are affected by host age. Older mice harboured greater numbers of the parasites’ offspring than younger mice. Protective immune responses that did not vary with age were dominated by HSPC, while ageing specifically eroded adaptive immunity, with reduced numbers of naïve T cells, poor T cell responsiveness to parasites, and impaired antibody production. We identified immune factors consistent with previously-reported immune responses to helminths, and also revealed novel interactions between helminths and HSPC maturation. Our approach thus allowed disentangling the concurrent effects of ageing and infection across the full maturation cycle of the immune response and highlights the potential of such approaches to improve understanding of the immune system within the whole organism.

Common methods of biological age estimation

At present, no single indicator could be used as a golden index to estimate aging process. The biological age (BA), which combines several important biomarkers with mathematical modeling, has been proposed for >50 years as an aging estimation method to replace chronological age (CA). The common methods used for BA estimation include the multiple linear regression (MLR), the principal component analysis (PCA), the Hochschild's method, and the Klemera and Doubal's method (KDM). The fundamental differences in these four methods are the roles of CA and the selection criteria of aging biomarkers. In MLR and PCA, CA is treated as the selection criterion and an independent index. The Hochschild's method and KDM share a similar concept, making CA an independent variable. Previous studies have either simply constructed the BA model by one or compared the four methods together. However, reviews have yet to illustrate and compare the four methods systematically. Since the BA model is a potential estimation of aging for clinical use, such as predicting onset and prognosis of diseases, improving the elderly's living qualities, and realizing successful aging, here we summarize previous BA studies, illustrate the basic statistical steps, and thoroughly discuss the comparisons among the four common BA estimation methods.

Changes in behaviors of male C57BL/6J mice across adult life span and effects of dietary restriction

Behavioral analysis is a high-end read-out of aging impact on an organism, and here, we have analyzed behaviors in 4-, 22-, and 28-month-old male C57BL/6J with a broad range of tests. For comparison, a group of 28-month-old males maintained on dietary restriction (DR) was included. The most conspicuous alteration was the decline in exploration activity with advancing age. Aging also affected other behaviors such as motor skill acquisition and grip strength, in contrast to latency to thermal stimuli and visual placement which were unchanged. Object recognition tests revealed intact working memory at 28 months while memory recollection was impaired already at 22 months. Comparison with female C57BL/6J (Fahlström et al., Neurobiol Aging 32:1868-1880, 2011) revealed that alterations in aged males and females are similar and that several of the behavioral indices correlate with age in both sexes. Moreover, we examined if behavioral indices in 22-month-old males could predict remaining life span as suggested in the study by Ingram and Reynolds (Exp Aging Res 12(3):155-162, 1986) and found that exploratory activity and motor skills accounted for up to 65% of the variance. Consistent with that a high level of exploratory activity and preserved motor capacity indicated a long post-test survival, 28-month-old males maintained on DR were more successful in such tests than ad libitum fed age-matched males. In summary, aged C57BL/6J males are marked by a reduced exploratory activity, an alteration that DR impedes. In light of recently published data, we discuss if a diminishing drive to explore may associate with aging-related impairment of central aminergic pathways.

Computational biology for ageing

High-throughput genomic and proteomic technologies have generated a wealth of publicly available data on ageing. Easy access to these data, and their computational analysis, is of great importance in order to pinpoint the causes and effects of ageing. Here, we provide a description of the existing databases and computational tools on ageing that are available for researchers. We also describe the computational approaches to data interpretation in the field of ageing including gene expression, comparative and pathway analyses, and highlight the challenges for future developments. We review recent biological insights gained from applying bioinformatics methods to analyse and interpret ageing data in different organisms, tissues and conditions.

Can rodent longevity studies be both short and powerful?

Many rodent experiments have assessed effects of diets, drugs, genes, and other factors on life span. A challenge with such experiments is their long duration, typically over 3.5 years given rodent life spans, thus requiring significant time costs until answers are obtained. We collected longevity data from 15 rodent studies and artificially truncated them at 2 years to assess the extent to which one will obtain the same answer regarding mortality effects. When truncated, the point estimates were not significantly different in any study, implying that in most cases, truncated studies yield similar estimates. The median ratio of variances of coefficients for truncated to full-length studies was 3.4, implying that truncated studies with roughly 3.4 times as many rodents will often have equivalent or greater power. Cost calculations suggest that shorter studies will be more expensive but perhaps not so much to not be worth the reduced time.

Indicators of "healthy aging" in older women (65-69 years of age). A data-mining approach based on prediction of long-term survival

Background

Prediction of long-term survival in healthy adults requires recognition of features that serve as early indicators of successful aging. The aims of this study were to identify predictors of long-term survival in older women and to develop a multivariable model based upon longitudinal data from the Study of Osteoporotic Fractures (SOF).

Methods

We considered only the youngest subjects (n = 4,097) enrolled in the SOF cohort (65 to 69 years of age) and excluded older SOF subjects more likely to exhibit a "frail" phenotype. A total of 377 phenotypic measures were screened to determine which were of most value for prediction of long-term (19-year) survival. Prognostic capacity of individual predictors, and combinations of predictors, was evaluated using a cross-validation criterion with prediction accuracy assessed according to time-specific AUC statistics.

Results

Visual contrast sensitivity score was among the top 5 individual predictors relative to all 377 variables evaluated (mean AUC = 0.570). A 13-variable model with strong predictive performance was generated using a forward search strategy (mean AUC = 0.673). Variables within this model included a measure of physical function, smoking and diabetes status, self-reported health, contrast sensitivity, and functional status indices reflecting cumulative number of daily living impairments (HR ≥ 0.879 or RH ≤ 1.131; P < 0.001). We evaluated this model and show that it predicts long-term survival among subjects assigned differing causes of death (e.g., cancer, cardiovascular disease; P < 0.01). For an average follow-up time of 20 years, output from the model was associated with multiple outcomes among survivors, such as tests of cognitive function, geriatric depression, number of daily living impairments and grip strength (P < 0.03).

Conclusions

The multivariate model we developed characterizes a "healthy aging" phenotype based upon an integration of measures that together reflect multiple dimensions of an aging adult (65-69 years of age). Age-sensitive components of this model may be of value as biomarkers in human studies that evaluate anti-aging interventions. Our methodology could be applied to data from other longitudinal cohorts to generalize these findings, identify additional predictors of long-term survival, and to further develop the "healthy aging" concept.

Longevity is not influenced by prenatal programming of body size

Insulin-like growth factor (IGF) signaling is essential for achieving optimal body size during fetal development, whereas, in the adult, IGFs are associated with aging and age-related diseases. However, it is unclear as to what extent lifespan is influenced by events that occur during development. Here, we provide direct evidence that the exceptional longevity of mice with altered IGF signaling is not linked to prenatal programming of body size. Mice null for pregnancy-associated plasma protein-A (PAPP-A), an IGF-binding protein proteinase that increases local IGF bioavailability, are 60-70% the size of their wild-type littermates at birth and have extended median and maximum lifespan of 30-40%. In this study, PAPP-A(-/-) mice whose body size was normalized during fetal development through disruption of IgfII imprinting did not lose their longevity advantage. Adult-specific moderation of IGF signaling through PAPP-A inhibition may present a unique opportunity to improve lifespan without affecting important aspects of early life physiology.

Identification of markers for imminent death in mice used in longevity and aging research

The goal of this study was to identify objective criteria that would reliably predict imminent death in aged mice. Male and female ICR mice (age, 8 mo) were subcutaneously implanted with an identification chip for remote measurement of body temperature. Mice then were weighed and monitored regularly until spontaneous death occurred or until euthanasia was administered for humane reasons. Clinical signs that signaled implementation of euthanasia included inability to walk, lack of response to manipulation, large or ulcerated tumors, seizures, and palpable hypothermia. In mice that died spontaneously, gradual weight loss was the most frequent and earliest sign of imminent death. Hypothermia developed during the 2 wk prior to death. Slow or labored breathing were observed in about half of the mice before death. A composite score of temperature x weight can be used to provide an objective benchmark to signal increased observation or euthanasia of individual mice. Such assessment may allow the collection of terminal tissue samples without markedly altering longevity data, although application of this criterion may not be appropriate for all studies of longevity. Timely euthanasia of mice based on validated markers of imminent death can allow implementation of endpoints that alleviate terminal distress in aged mice, may not significantly affect longevity data, and can permit timely collection of biologic samples.

Genes and gene expression modules associated with caloric restriction and aging in the laboratory mouse

BACKGROUND

Caloric restriction (CR) counters deleterious effects of aging and, for most mouse genotypes, increases mean and maximum lifespan. Previous analyses of microarray data have identified gene expression responses to CR that are shared among multiple mouse tissues, including the activation of anti-oxidant, tumor suppressor and anti-inflammatory pathways. These analyses have provided useful research directions, but have been restricted to a limited number of tissues, and have focused on individual genes, rather than whole-genome transcriptional networks. Furthermore, CR is thought to oppose age-associated gene expression patterns, but detailed statistical investigations of this hypothesis have not been carried out.

RESULTS

Systemic effects of CR and aging were identified by examining transcriptional responses to CR in 17 mouse tissue types, as well as responses to aging in 22 tissues. CR broadly induced the expression of genes known to inhibit oxidative stress (e.g., Mt1, Mt2), inflammation (e.g., Nfkbia, Timp3) and tumorigenesis (e.g., Txnip, Zbtb16). Additionally, a network-based investigation revealed that CR regulates a large co-expression module containing genes associated with the metabolism and splicing of mRNA (e.g., Cpsf6, Sfpq, Sfrs18). The effects of aging were, to a considerable degree, similar among groups of co-expressed genes. Age-related gene expression patterns characteristic of most mouse tissues were identified, including up regulation of granulin (Grn) and secreted phosphoprotein 1 (Spp1). The transcriptional association between CR and aging varied at different levels of analysis. With respect to gene subsets associated with certain biological processes (e.g., immunity and inflammation), CR opposed age-associated expression patterns. However, among all genes, global transcriptional effects of CR were only weakly related to those of aging.

CONCLUSION

The study of aging, and of interventions thought to combat aging, has much to gain from data-driven and unbiased genomic investigations. Expression patterns identified in this analysis characterize a generalized response of mammalian cells to CR and/or aging. These patterns may be of importance in determining effects of CR on overall lifespan, or as factors that underlie age-related disease. The association between CR and aging warrants further study, but most evidence indicates that CR does not induce a genome-wide "reversal" of age-associated gene expression patterns.